While wind energy has been reported as the fastest growing among different sources of renewable energy, two critical issues are how to make wind energy cost effective and how to integrate it into electricity grids properly. The ability to predict power generated by wind not only allows the most effective integration of wind power into electricity grid but also makes it possible to have an optimal maintenance scheduling that can reduce cost significantly. This research investigates the practical use of Self Organizing Map (SOM) as a special type of neural network based forecasting method. In this paper, forecasting the average, maximum and minimum of one-day-ahead wind speed based on the past wind speed states of the previous 24 hours is the objective.